arXiv subjects

Shiwen Zhang

Publications and source records attributed to Shiwen Zhang.

At least 19 recordsLinked to original sources

arXiv2026

One-Dimensional Divergence-Type Jacobi Operators Driven by The Doubling Map

We study one-dimensional Jacobi operators of divergence-gradient type on $\ell^2(\mathbb{Z}_{\geq 0})$, with coefficients generated by the doubling map, $a_n(x)=a(2^n x \mathrm{mod} 1)$. For continuous positive sampling functions, we show that the almost-sure essential spectrum is an interval containing the bottom of the spectrum. Under the assumption $a\in C^1(\mathbb{T})$, we prove square-root asymptotics for the integrated density of states and obtain a small-energy expansion for the Lyapunov exponent as $E\to0^+$. The leading term of the Lyapunov exponent is linear and positive precisely under a nondegeneracy condition on the sampling function. In this case, we prove a large-deviation estimate for the transfer matrices with an explicit rate depending on $E$. As consequences, we obtain local H\"older continuity of the Lyapunov exponent and the integrated density of states near the bottom of the spectrum, as well as the Anderson localization. Some of these results extend the small-coupling results of Chulaevsky--Spencer and Bourgain--Schlag for Schr\"odinger operators generated by the doubling map to divergence-gradient type operators. The arguments adapt methods from those works, but require uniform control in the small-energy parameter and refined correlation estimates reflecting the divergence-gradient structure.

math.SP↗

arXiv2026

TeleStyle V2: Beyond Content-Preserving Style Transfer with Self-Distillation and Distribution-Matching-Distillation

Given a content reference and a style reference, content-preserving style transfer requires the model to generate stylized outputs with content and style consistency. We introduced TeleStyle V1 to tackle this problem. However, TeleStyle V1 is trained with photorealistic content reference and artistic style reference, which makes it incapable to cope with artistic content reference and realistic style reference in most cases. In this paper, we designed a Self-Distillation data synthesis strategy to construct such triplets from TeleStyle V1. Trained with such self-distilled triplets, our TeleStyle V2 supports Content-Style references in the forms of Realistic-and-Realistic (RnR), Realistic-and-Stylized (RnS), Stylized-and-Realistic (SnR), Stylized-and-Stylized (SnS). In addition, we found Distribution Matching Distillation could preserve the general text-guided image editing capability of the foundation model and fix the content consistency degradation caused by SFT process. Through quantitative evaluations, our TeleStyleV2-QIE-2509-DMD performs at least on par with Qwen-Image-Edit-2509-DMD, demonstrating strong general image editing skills beyond content-preserving style transfer. We observed the content/style reference order confusion problem in TeleStyle V1 and further introduced prompt enhancer to solve it. TeleStyle V2 uses Qwen-Image-Edit's VLM encoder, Qwen2.5-VL-7B, to generate content prompt and style prompt for free. TeleStyle V2 could achieve comparable style transfer performance with state-of-the-art commercial model, gemini-3-pro-image-preview.

cs.CV↗

arXiv2026

On zero-sum polytopes: reciprocity, rigidity, and cyclic sieving

Let $G$ be a finite abelian group of order $n$, and let $\mathsf M(G,m)$ denote the set of zero-sum sequences over $G$ of length $m$. We introduce the zero-sum polytope $\mathcal P_G$, a rational polytope of dimension $n-1$, whose lattice points encode zero-sum sequences: \[ |\mathsf M(G,m)|=|m\mathcal P_G\cap \mathbb Z^n|. \] This naturally realizes the enumeration of zero-sum sequences as a problem in rational Ehrhart theory, which leads to a combinatorial reciprocity theorem identifying the negative evaluations of the corresponding counting quasipolynomial with zero-sum sequences of full support. Our main results establish a face-stratified rigidity for zero-sum polytopes: whenever two such polytopes have equal total lattice point counts at specific dilations, the dimension-wise open-face strata are equinumerous. Moreover, we study the natural $\operatorname{Aut}(G)$-action on $\mathcal P_G$, derive equivariant generating functions and reciprocity formulas, and obtain cyclic sieving phenomena for natural cyclic actions.

math.CO↗

arXiv2026

Style-CCL: Content-Preserving Style Transfer via Curriculum Continual Learning

Content-Preserving Style transfer, given content and style references, remains challenging for Diffusion Transformers (DiTs) due to entangled content and style features. With a reverse triplet synthesis pipeline to build a million-scale training set and a dual-branch Style-Content DiT (SC-DiT) that decouples style and content via separate ROPE embeddings and causal masking, we observe that such a one-stage training paradigm on mixed style categories causes semantic styles to dominate, hindering texture style learning, and harming content preservation. To address these issues, we propose Style-CCL, a Multi-Stage Curriculum Continual Learning framework that trains SC-DiT from semantic (easy) to texture (hard) styles, and from clean to synthetic data, with Random Memory Rehearsal across stages to avoid catastrophic forgetting. Extensive experiments demonstrate that our Style-CCL achieves state-of-the-art performance in three core metrics: style similarity, content consistency, and aesthetic quality.

cs.CV↗

arXiv2026

Edge localization and Lifshitz tails for graphs with Ahlfors regular volume growth

In this work, we study the Anderson model on graphs with Ahlfors $α$-regular volume growth. We show that, under mild regularity assumptions of the random distribution, Lifshitz-tail type estimates near the bottom of the spectrum lead to exponential decay of fractional moments of the Green's function and thus spectral and dynamical localization at low energies. This generalizes the result of [4] from the lattice $\mathbb{Z}^d$ to Ahlfors $α$-regular graphs. In addition, we establish Lifshitz tail estimates for the integrated density of states, with the Lifshitz exponent determined by the ratio of the volume growth rate and the random walk dimension of the underlying graphs, under certain assumptions on low lying eigenvalues of the Dirichlet and Neumann Laplacian on the graph. As an application, we verify all conditions on the Sierpinski gasket graph and obtain that, under mild regularity assumptions of the random distribution, for any fixed disorder, the Anderson model on the Sierpinski gasket graph has pure point spectrum and exhibits strong dynamical localization near the bottom of the spectrum.

math-ph↗

arXiv2026

LocalSUG: City-Preference-Enhanced LLM for Query Suggestion in Local-Life Services

In local-life service platforms, query suggestion reduces user effort by generating candidate queries from input prefixes. Traditional multi-stage systems rely heavily on historical popular queries, limiting their ability to capture long-tail and emerging demand. Although LLMs provide strong semantic generalization, their deployment in local-life services faces three challenges: insufficient city-preference awareness, exposure bias in preference optimization, and strict online latency constraints. We propose LocalSUG, an LLM-based query suggestion framework for local-life services. LocalSUG mines city-preference-enhanced candidates from term co-occurrence and injects them into prompts as dynamic references rather than fusing them into model parameters. This allows the model to adapt to changing city preferences, such as merchant openings or closures, while reducing stale or locally invalid suggestions. We further introduce a beam-search-driven GRPO algorithm to align training with inference-time decoding and optimize relevance together with business-oriented rewards. Finally, quality-aware beam acceleration and vocabulary pruning reduce online latency while preserving generation quality. Offline evaluations and large-scale online A/B testing show that LocalSUG improves CTR by +0.35% and reduces the no-result rate by 3.98%, demonstrating its effectiveness in real-world deployment.

cs.CL↗

arXiv2026

Tele-Omni: a Unified Multimodal Framework for Video Generation and Editing

Recent advances in diffusion-based video generation have substantially improved visual fidelity and temporal coherence. However, most existing approaches remain task-specific and rely primarily on textual instructions, limiting their ability to handle multimodal inputs, contextual references, and diverse video generation and editing scenarios within a unified framework. Moreover, many video editing methods depend on carefully engineered pipelines tailored to individual operations, which hinders scalability and composability. In this paper, we propose Tele-Omni, a unified multimodal framework for video generation and editing that follows multimodal instructions, including text, images, and reference videos, within a single model. Tele-Omni leverages pretrained multimodal large language models to parse heterogeneous instructions and infer structured generation or editing intents, while diffusion-based generators perform high-quality video synthesis conditioned on these structured signals. To enable joint training across heterogeneous video tasks, we introduce a task-aware data processing pipeline that unifies multimodal inputs into a structured instruction format while preserving task-specific constraints. Tele-Omni supports a wide range of video-centric tasks, including text-to-video generation, image-to-video generation, first-last-frame video generation, in-context video generation, and in-context video editing. By decoupling instruction parsing from video synthesis and combining it with task-aware data design, Tele-Omni achieves flexible multimodal control while maintaining strong temporal coherence and visual consistency. Experimental results demonstrate that Tele-Omni achieves competitive performance across multiple tasks.

cs.CV↗

arXiv2026

TeleStyle: Content-Preserving Style Transfer in Images and Videos

Content-preserving style transfer, generating stylized outputs based on content and style references, remains a significant challenge for Diffusion Transformers (DiTs) due to the inherent entanglement of content and style features in their internal representations. In this technical report, we present TeleStyle, a lightweight yet effective model for both image and video stylization. Built upon Qwen-Image-Edit, TeleStyle leverages the base model's robust capabilities in content preservation and style customization. To facilitate effective training, we curated a high-quality dataset of distinct specific styles and further synthesized triplets using thousands of diverse, in-the-wild style categories. We introduce a Curriculum Continual Learning framework to train TeleStyle on this hybrid dataset of clean (curated) and noisy (synthetic) triplets. This approach enables the model to generalize to unseen styles without compromising precise content fidelity. Additionally, we introduce a video-to-video stylization module to enhance temporal consistency and visual quality. TeleStyle achieves state-of-the-art performance across three core evaluation metrics: style similarity, content consistency, and aesthetic quality. Code and pre-trained models are available at https://github.com/Tele-AI/TeleStyle

cs.CV↗

arXiv2026

Upper and Lower Bounds for The Quantum Dynamics of One-Dimensional Divergence-Type Random Jacobi Operators

We study quantum transport for the discrete one-dimensional random Jacobi operator of divergence-gradient type. For strictly positive and bounded random variables, we analyze the q-moments of the position operator and establish both upper and lower power-law bounds on their growth. Our approach relies on the asymptotic behavior of the integrated density of states and the Lyapunov exponent near the critical energy 0, previously obtained by Pastur and Figotin. A key ingredient in our analysis is the large deviation-type estimates explored via the phase formalism, which play a central role in deriving bounds on the growth of the transfer matrices.

math-ph↗

arXiv2026

QwenStyle: Content-Preserving Style Transfer with Qwen-Image-Edit

Content-Preserving Style transfer, given content and style references, remains challenging for Diffusion Transformers (DiTs) due to its internal entangled content and style features. In this technical report, we propose the first content-preserving style transfer model trained on Qwen-Image-Edit, which activates Qwen-Image-Edit's strong content preservation and style customization capability. We collected and filtered high quality data of limited specific styles and synthesized triplets with thousands categories of style images in-the-wild. We introduce the Curriculum Continual Learning framework to train QwenStyle with such mixture of clean and noisy triplets, which enables QwenStyle to generalize to unseen styles without degradation of the precise content preservation capability. Our QwenStyle V1 achieves state-of-the-art performance in three core metrics: style similarity, content consistency, and aesthetic quality.

cs.CV↗

arXiv2025

TeleWorld: Towards Dynamic Multimodal Synthesis with a 4D World Model

World models aim to endow AI systems with the ability to represent, generate, and interact with dynamic environments in a coherent and temporally consistent manner. While recent video generation models have demonstrated impressive visual quality, they remain limited in real-time interaction, long-horizon consistency, and persistent memory of dynamic scenes, hindering their evolution into practical world models. In this report, we present TeleWorld, a real-time multimodal 4D world modeling framework that unifies video generation, dynamic scene reconstruction, and long-term world memory within a closed-loop system. TeleWorld introduces a novel generation-reconstruction-guidance paradigm, where generated video streams are continuously reconstructed into a dynamic 4D spatio-temporal representation, which in turn guides subsequent generation to maintain spatial, temporal, and physical consistency. To support long-horizon generation with low latency, we employ an autoregressive diffusion-based video model enhanced with Macro-from-Micro Planning (MMPL)--a hierarchical planning method that reduces error accumulation from frame-level to segment-level-alongside efficient Distribution Matching Distillation (DMD), enabling real-time synthesis under practical computational budgets. Our approach achieves seamless integration of dynamic object modeling and static scene representation within a unified 4D framework, advancing world models toward practical, interactive, and computationally accessible systems. Extensive experiments demonstrate that TeleWorld achieves strong performance in both static and dynamic world understanding, long-term consistency, and real-time generation efficiency, positioning it as a practical step toward interactive, memory-enabled world models for multimodal generation and embodied intelligence.

cs.CV↗

arXiv2025

DreamO: A Unified Framework for Image Customization

Recently, extensive research on image customization (e.g., identity, subject, style, background, etc.) demonstrates strong customization capabilities in large-scale generative models. However, most approaches are designed for specific tasks, restricting their generalizability to combine different types of condition. Developing a unified framework for image customization remains an open challenge. In this paper, we present DreamO, an image customization framework designed to support a wide range of tasks while facilitating seamless integration of multiple conditions. Specifically, DreamO utilizes a diffusion transformer (DiT) framework to uniformly process input of different types. During training, we construct a large-scale training dataset that includes various customization tasks, and we introduce a feature routing constraint to facilitate the precise querying of relevant information from reference images. Additionally, we design a placeholder strategy that associates specific placeholders with conditions at particular positions, enabling control over the placement of conditions in the generated results. Moreover, we employ a progressive training strategy consisting of three stages: an initial stage focused on simple tasks with limited data to establish baseline consistency, a full-scale training stage to comprehensively enhance the customization capabilities, and a final quality alignment stage to correct quality biases introduced by low-quality data. Extensive experiments demonstrate that the proposed DreamO can effectively perform various image customization tasks with high quality and flexibly integrate different types of control conditions.

cs.CV↗

arXiv2025

Beyond the Surface: A Solution-Aware Retrieval Model for Competition-level Code Generation

In competitive programming task, problem statements are often embedded within elaborate narrative backgrounds, requiring deep understanding of the underlying solutions to successfully complete the tasks. Current code generation models primarily focus on token-level semantic modeling, highly susceptible to distractions from irrelevant narrative statements. Inspired by RAG, retrieving reference code with similar solutions may help enhance model performance on difficult problems. However, existing retrieval models also emphasize surface-level semantic similarity, neglecting the deeper solution-level logical similarities that are critical in competitive programming. Therefore, designing ranking models capable of accurately identifying and retrieving problems and corresponding codes remains an urgent research problem in competitive code generation. In this paper, we propose SolveRank, a solution-aware ranking model empowered by synthetic data for competitive programming tasks. Specifically, we leverage the DeepSeek-R1 model to generate logically equivalent but differently phrased new problems, verified by GPT-4o for solution consistency. Then, we train SolveRank with these as positive samples and BM25/random-retrieved problems as negatives. During inference, SolveRank retrieves relevant problems and corresponding code from the corpus to assist a downstream code generator. Experiments on the xCodeEval dataset demonstrate that SolveRank outperforms SOTA ranking methods in precision and recall metrics, and boosts code generation performance for difficult problems.

cs.IR↗

arXiv2025

AviationLLM: An LLM-based Knowledge System for Aviation Training

Aviation training is a core link in ensuring flight safety, improving industry efficiency and promoting sustainable development. It not only involves flight simulation but also requires the learning of a great deal of professional aviation theory knowledge. In the existing training system, the knowledge is mainly imparted by the the instructors. However, the number of instructors is limited and the professional answers obtained from the Internet are not accurate enough, resulting in low training efficiency. To address this, we introduced LLM, but the basic pre-trained model cannot provide accurate answers to professional fields, so we fine-tuned it. Traditional Supervised Fine-Tuning (SFT) risk generating superficially plausible but factually incorrect responses due to insufficient data coverage. To address this, we employ Direct Preference Optimization(DPO). This paper proposes Retrieval-Augmented LLM Alignment via Direct Preference Optimization(RALA-DPO). We select open source pre-trained LLM Qwen and adapt it to aviation theory training through DPO-based domain alignment. Simultaneously, to mitigate hallucinations caused by training data biases, knowledge obsolescence, or domain knowledge gaps, we implement Retrieval-Augmented Generation(RAG) technology that combines generative and retrieval models. RALA-DPO effectively retrieves relevant information from external knowledge bases and delivers precise and high-quality responses through the generative model. Experimental results demonstrate that RALA-DPO can improve accuracy in response to professional aviation knowledge. With integrated RAG mechanisms, this system can further improve the accuracy of answers and achieve zero-cost knowledge updates simultaneously.

cs.AI↗

arXiv2025

CDST: Color Disentangled Style Transfer for Universal Style Reference Customization

We introduce Color Disentangled Style Transfer (CDST), a novel and efficient two-stream style transfer training paradigm which completely isolates color from style and forces the style stream to be color-blinded. With one same model, CDST unlocks universal style transfer capabilities in a tuning-free manner during inference. Especially, the characteristics-preserved style transfer with style and content references is solved in the tuning-free way for the first time. CDST significantly improves the style similarity by multi-feature image embeddings compression and preserves strong editing capability via our new CDST style definition inspired by Diffusion UNet disentanglement law. By conducting thorough qualitative and quantitative experiments and human evaluations, we demonstrate that CDST achieves state-of-the-art results on various style transfer tasks.

cs.CV↗

arXiv2025

ContribChain: A Stress-Balanced Blockchain Sharding Protocol with Node Contribution Awareness

Existing blockchain sharding protocols have focused on eliminating imbalanced workload distributions. However, even with workload balance, disparities in processing capabilities can lead to differential stress among shards, resulting in transaction backlogs in certain shards. Therefore, achieving stress balance among shards in the dynamic and heterogeneous environment presents a significant challenge of blockchain sharding. In this paper, we propose ContribChain, a blockchain sharding protocol that can automatically be aware of node contributions to achieve stress balance. We calculate node contribution values based on the historical behavior to evaluate the performance and security of nodes. Furthermore, we propose node allocation algorithm NACV and account allocation algorithm P-Louvain, which both match shard performance with workload to achieve stress balance. Finally, we conduct extensive experiments to compare our work with state-of-the-art baselines based on real Ethereum transactions. The evaluation results show that P-Louvain reduces allocation execution time by 86% and the cross-shard transaction ratio by 7.5%. Meanwhile, ContribChain improves throughput by 35.8% and reduces the cross-shard transaction ratio by 16%.

cs.NI↗

arXiv2024

Spectrum and Lifshitz tails for the Anderson model on the Sierpinski gasket graph

In this work, we study the Anderson model on the Sierpinski gasket graph. We first identify the almost sure spectrum of the Anderson model when the support of the random potential has no gaps. We then prove the existence of the integrated density states of the Anderson model and show that it has Lifshitz tails with Lifshitz exponent determined by the ratio of the volume growth rate and the random walk dimension of the Sierpinski gasket graph.

math-ph↗

arXiv2024

Hyper-parameter tuning for text guided image editing

The test-time finetuning text-guided image editing method, Forgedit, is capable of tackling general and complex image editing problems given only the input image itself and the target text prompt. During finetuning stage, using the same set of finetuning hyper-paramters every time for every given image, Forgedit remembers and understands the input image in 30 seconds. During editing stage, the workflow of Forgedit might seem complicated. However, in fact, the editing process of Forgedit is not more complex than previous SOTA Imagic, yet completely solves the overfitting problem of Imagic. In this paper, we will elaborate the workflow of Forgedit editing stage with examples. We will show how to tune the hyper-parameters in an efficient way to obtain ideal editing results.

cs.CV↗